update server for argsparser

huangmingming

2023-01-29 c4d60d9a237d97d68db352d5b3e3be9c332d0d79

update server for argsparser

 funasr/runtime/python/grpc/grpc_main_server.py

@@ -8,7 +8,7 @@
      server = grpc.server(futures.ThreadPoolExecutor(max_workers=10),
                        # interceptors=(AuthInterceptor('Bearer mysecrettoken'),)
                           )
      paraformer_pb2_grpc.add_ASRServicer_to_server(ASRServicer(args.user_allowed), server)
      paraformer_pb2_grpc.add_ASRServicer_to_server(ASRServicer(args.user_allowed, args.model, args.sample_rate), server)
      port = "[::]:" + str(args.port)
      server.add_insecure_port(port)
      server.start()
@@ -21,10 +21,22 @@
                        type=int,
                        default=10095,
                        help="grpc server port")
                        
    parser.add_argument("--user_allowed",
                        type=str,
                        default="project1_user1|project1_user2|project2_user3",
                        help="allowed user for grpc client")
                        
    parser.add_argument("--model",
                        type=str,
                        default="damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch",
                        help="model from modelscope")
                        
    parser.add_argument("--sample_rate",
                        type=int,
                        default=16000,
                        help="audio sample_rate from client")    
                        


    args = parser.parse_args()

 funasr/runtime/python/grpc/grpc_server.py

@@ -5,26 +5,20 @@
import paraformer_pb2_grpc
import time


from paraformer_pb2 import Response


from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks

inference_16k_pipline = pipeline(
   task=Tasks.auto_speech_recognition,
   model='damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8358-tensorflow1')

auth_user = ['zksz_futureTV_1','zksz_dangaomei_1','zksz_test_1','zksz_test_2','zksz_test_3','zksz_test_4','zksz_test_5','zksz_test_6','zksz_test_7']


class ASRServicer(paraformer_pb2_grpc.ASRServicer):
    def __init__(self):
    def __init__(self, user_allowed, model, sample_rate):
        print("ASRServicer init")
        self.init_flag = 0
        self.client_buffers = {}
        self.client_transcription = {}
        self.auth_user = user_allowed.split("|")
        self.inference_16k_pipline = pipeline(task=Tasks.auto_speech_recognition, model=model)
        self.sample_rate = sample_rate

    def clear_states(self, user):
        self.clear_buffers(user)
@@ -46,13 +40,13 @@
        
            
        for req in request_iterator:
            if req.user not in auth_user:
            if req.user not in self.auth_user:
                result = {}
                result["success"] = False
                result["detail"] = "Not Authorized user: %s " % req.user
                result["text"] = ""
                yield Response(sentence=json.dumps(result), user=req.user, action="terminate", language=req.language)
            if req.isEnd: #end grpc
            elif req.isEnd: #end grpc
                print("asr end")
                self.disconnect(req.user)
                result = {}
@@ -80,7 +74,7 @@
                    yield Response(sentence=json.dumps(result), user=req.user, action="waiting", language=req.language)
                else:
                    begin_time = int(round(time.time() * 1000))
                    tmp_data = self.client_buffers[req.user] #TODO make a test, about local variable in class parralle circumstance.
                    tmp_data = self.client_buffers[req.user]
                    self.clear_states(req.user)
                    result = {}
                    result["success"] = True
@@ -98,7 +92,7 @@
                        print ("user: %s , delay(ms): %s, error: %s " % (req.user, delay_str, "data_is_not_long_enough"))
                        yield Response(sentence=json.dumps(result), user=req.user, action="finish", language=req.language)
                    else:                           
                        asr_result = inference_16k_pipline(audio_in=tmp_data, audio_fs = 16000)
                        asr_result = self.inference_16k_pipline(audio_in=tmp_data, audio_fs = self.sample_rate)
                        if "text" in asr_result:
                            asr_result = asr_result['text']
                        else: